Domain Adaptation and Few-Shot Learning

Removing Inter-Experimental Variability from Functional Data in Systems Neuroscience

This paper develops a domain adaptation framework to remove inter-experimental variability in systems neuroscience data while preserving biological signals, showing improved performance over existing methods.

A Biologically Plausible Model for Continual Learning using Synaptic Weight Attractors

This paper introduces a biologically plausible method for continual learning in ANNs that prevents catastrophic forgetting and promotes synaptic multi-modality.

The Geometry of Concept Learning

The paper proposes a geometric framework for few-shot learning using neural manifolds and a single plastic neuron, demonstrating its effectiveness on visual concepts and linking neurophysiology to behavior.

Leveraging prior concept learning improves ability to generalize from few examples in computational models of human object recognition

This paper shows that reusing high-level visual features improves deep learning models' ability to learn from few examples, mimicking human visual concept learning.

Deep Learning-Based Unlearning of Dataset Bias for MRI Harmonisation and Confound Removal

This paper introduces a deep learning framework to reduce scanner bias and other confounds in MRI data, improving harmonization and generalizability across datasets.

DISSECT: DISentangle SharablE ConTent for Multimodal Integration and Crosswise-mapping

The paper introduces a method to separate domain-specific information using a cross-domain autoencoder and orthogonal gate block for improved domain translation.

Unsupervised domain adaptation for the automated segmentation of neuroanatomy in MRI: a deep learning approach

This paper develops an unsupervised domain adaptation method to improve automated neuroanatomy segmentation in MRI across different domains without target domain annotations.

Evaluation of domain adaptation approaches for robust classification of heterogeneous biological data sets

This study evaluates domain adaptation methods to improve classification accuracy in heterogeneous biological data, finding that EasyAdapt outperforms other techniques.

Improving Whole-Brain Neural Decoding of fMRI with Domain Adaptation

This paper improves whole-brain fMRI neural decoding using domain adaptation, achieving an 8.94% accuracy improvement by leveraging pre-existing source data.

Focused learning promotes continual task performance in humans

This study shows that focused training improves continual task performance in humans by enabling factorised task representations, which reduces interference between tasks.

SupportNet: a novel incremental learning framework through deep learning and support data

SupportNet addresses catastrophic forgetting in incremental learning by combining deep learning with SVM to retain old knowledge while learning new data, outperforming existing methods in multiple tasks.